Samantha LLM
Samantha LLM is a persistent memory and persona system for agentic LLMs. It gives your AI coding assistant long-term memory, project awareness, and a consistent personality across sessions — so you never have to re-explain context, decisions, or preferences.
The Problem
Every time you start a new session with an AI coding assistant, you start from zero. It doesn’t know what you worked on yesterday, what decisions you made last week, or how you like to work. You waste time re-explaining the same context over and over.
Samantha LLM fixes this:
- Persistent memory — Your assistant remembers what you worked on, what you decided, and why. Across every session.
- Automatic — A background system analyzes each session and creates structured memories. No manual note-taking.
- Project-aware — Active projects, status, and context are tracked and surfaced automatically. Pick up right where you left off.
- Learns your preferences — Tell your assistant something once and it sticks. Environment details, coding style, tooling choices — all retained.
Quickstart
# Clone and install
git clone git@github.com:criswell/samantha-llm.git
cd samantha-llm
./samantha-llm install
# Set up your agent (Claude Code, GitHub Copilot, Abacus.ai, or custom)
samantha-llm setup
# Go to your project and start
cd /path/to/your/project
samantha-llm start
On first launch, Samantha will have no memories. Describe what you want to work on and she’ll build context that persists across future sessions.
Features
- Memory System — Short-term, long-term, and procedural memory that builds over time
- Subconscious Processing — Automatic background analysis of sessions into structured memories
- Project Management — Track active projects, status, and context across sessions
- Custom Personas — Use the default Samantha Hartwell persona or create your own
- Memory Search — Semantic search across all stored memories
- Ralph Mode — Iterative coding framework for complex tasks with autonomous work-test-evaluate loops
- Multi-Agent Support — Configure multiple LLM tools and switch between them
Supported Agents
- Claude Code
- GitHub Copilot CLI
- Abacus.ai Deep Agent
- Any agentic LLM that accepts a text prompt